From: Osher Doctorow osher.domain.name.hidden, Fri. Sept. 6, 2002 8:36AM
After my discouragement of yesterday, I have decided to give myself one more
chance to try to be compatible with everything-list. I have just
downloaded J. Schmidthuber's *A computer scientist's view of life, the
universe, and everything,* (1997), and it is well enough written that I
apparently will be able to understand it. I also have one of the other two
papers that were available on the everything-list site, which I understood
fairly well several weeks ago.
I will take this opportunity to write/type a few words about Knowledge (K
for short) rather than information (I), distinguishing between them as the
semantic (meaning) part of that whose syntactic part is information vs
information itself.
There are several directions in which I have developed K, but the simplest
way is to consider that K contains primitive pointlike or stringlike
elements which may have fuzzy truth values (on a scale between 0 and 1 for
the non-trivial cases) or probabilities or both assigned to them. Let us
call these K-points for brevity. Each K-point has MEANING, and I will
regard this as a primitive undefined concept in this presentation, although
one can develop things from several other viewpoints. The word *MEANING*,
however, is to be used in practice rather similarly to its dictionary
definition(s) and intuitively is like ideas, thoughts, cognitions, provided
that they are accompanied by *understanding* rather than merely regarded as
sounds or sights or perceptions with nothing that can be specified behind
them. I do relate it here at all to the computational linguistics idea of
*meaning*.
Knowledge (K) is regarded as continuous or piecewise continuous and
connected or piecewise connected, and could theoretically either increase or
decrease or remain constant, although in fact I postulate somewhat
analogously although apparently not structurally related to entropy in
thermodynamics that K increases in time in the universe. In fact, letting
E symbolize entropy, I postulate that the rate of increase of K in time
exceeds the rate of increase of E in time, symbolically:
1) Dt(K - E) > 0
where Dt is the partial derivative with respect to time, although I am open
to generalizing it to covariant or gauge derivatives and so on. Equation
(1) has an interesting interpretation, namely, that instead of disorder
increasing overall in the universe with time, the ordered part of Knowledge
increases with time - possibly by matter converting to radiation in whole or
in part, or possibly by some other scenario. Even the notion of a
radiation form of life is not excluded in these considerations - in fact, it
may be indicated. It might be in some places combined with material form
of life, as in the human brain, where the global aspects may relate more to
radiation and the local aspects more to matter. There is a well accepted
physical theory of the initial radiation-dominated era of the universe
succeeded by a matter-dominated era in which radiation still plays an
important part, and several theorists consider that a radiation type of era
will eventually constitute a third era.
Does this mean that digital computes do not do anything? No. They
calculate very fast. They store discrete steps and discrete Knowledge
representations (or attempted representations via syntax) and discrete
syntax. When they calculate very fast, they sometimes produce numerical
approximations to solutions of differential or integral equations which we
do not know how to produce otherwise, and this helps increase Knowledge,
although I think that is it qualitatively somewhat inferior to CAUSAL
KNOWLEDGE. There is factual knowledge (details) about the real world,
there is causal knowledge about what causes or influences what in the real
world, and there is speculative or even fictional or fantasy *knowledge*
about things or events or processes that are not considered to be real or to
have real analogs in the real physical or even psychological worlds.
Digital computers can help factual knowledge, but so far they have not
helped causal knowledge much.
Does K (Knowledge) relate to multiple universes, multiple histories, etc.?
This is a more advanced question than I can deal with here. I think that
multiple universes and multiple histories are interesting ideas, but that at
the present time their logical and physical and philosophical structures
have not been well established. If they exist, then I have no doubt
intuitively that K applies to them as well.
Finally, the mathematical formulation of Causal Knowledge in my Rare Event
Theory (RET) resides in fuzzy multivalued logical x-->y or its
probability-statistics analogs or proximity function - geometry-topology
analogs. There are 3 types of x-->y, which correspond respectively to
Rare Events (Lukaciewicz and Rational Pavelka fuzzy multivalued logics (FML)
in the non-trivial case), Fairly Frequent Events (Product/Goguen FML), and
Very Frequent Events (Godel FML). Here Events is short for Events and/or
Processes. I spent about 23 years showing the correspondence here since
1980 with my wife Marleen. Some of it is in my paper in B. N. Kursunuglu
et al (2000), Kluwer Academic: N.Y. 89-97, and references therein.
Osher Doctorow
Received on Fri Sep 06 2002 - 09:27:31 PDT
This archive was generated by hypermail 2.3.0
: Fri Feb 16 2018 - 13:20:07 PST